General comments

This paper compares the efficiency of a Convection-Permitting Model (CPM : ICON-CLM 2.6.4 at 3km) and its driving Regional Climate Model (RCM : ICON-CLM 2.6.4 at 11km) to simulate some climate variables and discharge when forcing WaSim distributed hydrological model. The study is conducted over a cluster of catchments of small and medium size in central eastern Germany. The evaluated climate variables are the one relevant for hydrological modelling such as surface temperature, relative humidity, wind speed, global radiation and precipitation. The CPM exhibits no clear added-value for the simulation of all the climate variables. For precipitation, especially extreme precipitation, the increase in model resolution switches from a negative bias with the RCM to a strong positive bias and sometimes unrealistic values with the CPM. This behavior leads to an overestimation of discharge over the studied catchments even if the strongest recorded flood of the period remains underestimated.

Due to the scarcity of hydroclimatic studies using high resolution regional climate models, this paper is welcomed in the community, even if it consists of a simple case study using only one CPM-RCM couple and one hydrological model on one particular cluster of catchments. The study adds further scientific content in the CPM bibliography by assessing a CPM potential benefits and understanding their transferability to the hydrology. 

The article is interesting, well organized and globally well written. However, some important information is lacking and some important aspects need to be addressed to improve either the clarity of the message and the consistency of the paper to make it suitable for publication in HESS.

1 - According to the paper title, the aim of the study is the assessment of the added value of the CPM compared to the RCM on hydroclimatic simulations, meaning climate and related hydrological simulations. The evaluation of global radiation, wind speed and relative humidity, even if bringing interest, are too detailed for the real purpose of the paper. The evaluation results presented for those variables are discussed separately and are not connected in any way to the final output that is the hydrological simulations.

If keeping this level of details (annual cycles of global radiation, diurnal cycles of wind speed, relative humidity) I would expect to see :

• a) A quick comparison of Potential Evapotranspiration (PE) obtained from simulations and observations to assess the impact of these variables on hydrological model forcing.  PE could be directly accessible from the hydrological model output. If not, you could compute it with the Penman-Monteith equation from the climate data. PE annual and diurnal cycles would be at least as interesting and relevant for your study as diurnal cycles of wind speed and annual cycles of global radiation.
  
  
  
  
• b) If possible, the impact on these variables on snowmelt in the model, and their contribution to discharge biases.

2 - Some important results of the paper are discussed but never presented in tables or figures. On the contrary some others are, in my opinion, secondary if the analysis proposed in the previous comment is not produced. These intermediate results take up space at the detriment of the hydrological analysis which is the main point of the study.

• a) l308-311 and 426-428 : You present and discuss results concerning discharges above the 99.5th percentile but these results are never shown in tables or figures. Yet, one of the main reasons for CPM development and their use is to enhance impact studies for extreme events such as floods. Here, the results about extreme discharges are really important to show, either under the form of CDF comparison, QQ plots or even a table comparing biases for particular quantiles of discharge.
  
  
  
  
• b) Too much attention is brought to wind speed frequency evaluation, for really tiny frequency biases that probably do not affect discharge modelling at all. In my opinion, seeing intensity biases on the whole distributions would be more informative.  Figure 4 should be moved to supportive information and potentially replaced by a QQ plot of the type shown in figure 6.
  
  
  
  
• c) l221 - 225 : The main finding in Figure 6 (and later Figure 8)  that stands out is the significant overestimation of precipitation for ICON3km, with unrealistic values (> 100mm in 1 hour). Either here or in the discussion, these surprising results are not highlighted, nor discussed enough. What are the synoptic scale conditions leading to the occurrence of these values ? Where do they occur ? Are they numerical errors ? Are these findings in line with previous versions of ICON3km ?
  
  Furthermore, the median of precipitation above the 99.5th percentile is not representative of extreme values. An option is to conduct the same computation over a different quantiles (25th percentile, median, 90th percentile, 95th percentile and 99th percentile) of wet hourly rainfall (> 0.1mm/h or >1mm/h) distribution. The biases of ICON3km seem to exceed 200% for the highest quantiles of wet hourly precipitation.

3 - Figure 1 is a map showing the study area and its contrasted topography. In the rest of the text, the biases of climate variables are presented either aggregated on the whole area or without distinction of geographical location. As a case study over a limited area, it would be very interesting to use the distributed nature of the data to visualize certain results in a map format. For example, a map of extreme precipitation biases for the two climate models, even in their native resolution, would be of great help to understand the spatial distribution of these biases and their transfer role in hydrological biases. It could as well help to understand the spatial patterns of hourly extreme rainfall values (> 100mm/h, visible in figure 6). Are these values simulated on high elevation areas (potentially related to biases of the orography forcing) or randomly over the study area (numerical errors) ? 

This map could be included in the main body or in the supportive information depending on the relevance and of the article available space.

4 - The methodology section should be complemented by :

• a) A small paragraph presenting the hydro-climatic conditions of the study area (climate type, hydrological regime…)
  
  
  
  
• b) A new paragraph presenting hydrological model calibration and validation periods and choices. We understand later on in the results (l 269-270 and l275-276) that the authors chose to calibrate WaSim on the June 2013 flood and validate it over a winter flood, but no periods or dates are specified. This methodology remains therefore unclear and arises some questions:
  • The analysis of the paper focuses on the whole range of discharges, but not particularly on extremes (except for one particular flood). Why have you chosen to calibrate the hydrological model only on extreme events?
  • What are the exact period dates for calibration and validation?
  • Can certain hydrological processes, such as those governing long-term soil moisture, be excluded from calibration and therefore poorly simulated?
  • A classical modelling approach is to perform a split-sample test to calibrate and validate your hydrological model (KLEMEŠ, 1986). Given the short period of data, another recommended approach is to calibrate over the entire length of data (Arsenault et al., 2018; Shen et al., 2022). Have you tried one or the other approach? If so, it would be interesting to see a summary of the calibration/validation results. If not, could you justify your calibration choices?

The paragraph should clarify this aspect of methodology, by justifying the authors’ choices and answering the question above as best as possible.

Specific comments 

• In my opinion, the passive form is overused and makes the text sometimes difficult to read. The expression “was/were found to” is too frequent (lines 154, 170,  175, 188, 192, 193, 204, 211, 250, 258, 260, 274, 276, 283, 297, 306, 332, 341, 346, 358, 366, 369, 378, 392, 396, 403) and could be changed in many cases by a more active form. For example :  “ICON3km was found to overestimate” could be replaced by “ICON3km tends to…”, “ICON3km shows an overestimation” or “ICON3km overestimates…”, or “We notice an overestimation…”
• Table 1 : The table 1 in Introduction listing existing hydroclimatic studies using CPM is greatly appreciated, but needs to be completed with the most recent work up to date (Dale and Shelton, 2025; Xie et al., 2025, maybe others…). 
• l47: This sentence is misleading. The part of the sentence “come to offer substantial added value for flood simulation” implies that a consensus has been reached about the added-value of CPM on discharge modelling, which is not the case yet, thus justifying the interest of your study. The listed studies in Table 1 focus on various aspects of hydrological modelling, from low flows to floods, covering very different regions and climates and using diverse CPM and are too few to conclude. I would recommend the authors to take into account this comment and change this part of the sentence. 
• l81 : To compare point-based climate data to stations, are you applying a correction for the temperature vertical profile depending on grid mean elevation and station elevation ?
• l111: If I understand this part of methodology well, precipitation observations have been upscaled to the climate model resolution, only with an aggregation of the initial grid and no regridding ? No interpolation has been done ?
• l113-115 : This part of the sentence has to be rephrased to improve clarity.
• l125 : Can you justify the differences of duration increments between ICON3km and ICON11km ?
• In section 2.4 or 3.1, please explicitly detail the extent of the evaluation domain. Is it the rectangle displayed in figure 1.b or the mask of the catchments ?
• l146-147 and 148-149 : These results (temperature estimates and frequency distribution) are not linked to any figure or table. Please show them.
• l154 : Could you be more specific about the sign of the error (negative) ?
• l171-173 : This sentence should be rephrased to clarify the message : the passive voice and all the commas make it difficult to read.
• l176 : Can you convert this 2.5J/cm² bias reduction in percent ? It is hard to realize if it is a big improvement or negligible.
• l194-195 : This sentence is hardly understandable. I propose this modification : “The largest difference in the average monthly errors between the two models over the year is only 1.3%, …”. A figure could help visualize this aspect.
• l201-202 :  What is the measurement uncertainty for relative humidity ?
• L237 : change “In keeping is” by “We notice”
• A presentation and an analysis of the diurnal cycle of precipitation would be interesting, considering the task is done for global radiation and wind speed. In addition, it could help to understand the representation of summer convective precipitation by the model and the biases shown for July in Fig S4.. Depending on the length of the paper, this could fit in Supportive Information.
• L254 - 256 : This sentence should be rephrased. Here is a suggestion : “Additionally, the finer resolution of ICON3km improves the delineation of heavy rainfall events, preventing runoff generation and routing outside of the concerned catchments, as it occurs in RCM” 
• l259 : I suggest to cut the sentence after ICON3km and to start after it with : “This is also reflected…”
• l276-277 : Could you share an example of this behavior ? 
• l278 : “were too high” → “underestimates”
• l291 : The sentence of fig 10 legend is too long. Could you separate it into different parts ?
• l304 : What do you call the “full range of ICON11km meteorological data” ? 
• l306 : Why are you considering only catchments relative to the main stem ?
• l314 : In methodology, it is stated that the calibration was performed on the July 2013 flood. How did you calibrate the hydrological model on other catchments if discharge measurements are not available ? 
• l336-340 : This discussion is interesting. This behavior could have been checked easily in your study by looking if the temperature biases are more important on Tmin than Tmax. 
• l342-344 : The source of summer temperature improvement can be checked in your study by looking at the improvement of diurnal cycle of precipitation. Have you taken a look at it ?
• l352-353 : Wrong citation yea. It is Keller et al, 2016.
• Fig 4 and lines 366-367 in discussion : The underestimation of extremes is not clear. There is a very slight underrepresentation of very light winds (can we consider them extremes ?) but no visible results on strong winds. To state this, you should represent and analyze a QQ-plot of ICON3km and ICON11km against observations, as advised in Major comment 2.b, and change these lines.
• l380-381 : This sentence should be rephrased for clarity. 
• l403: I advice the authors to rephrase the sentence to something like : “ICON3km does not improve the simulation of monthly precipitation, exhibiting a negative bias as ICON11km”
• l414-415 and 425-426: You have to be careful in your statement here. The different studies are not analyzing the same aspects of hydrology. Some did for the whole discharge range and others only for floods. Please modify these statements accordingly.
• l436-437 : Complete the sentence here to remind that the study is a case study and the results are valid over these particular catchments with the specific used climate models. 
• Conclusion : a sentence should be added to contrast between the results and what is written line 47. Your study is not in line with recent studies and brings an interesting result : CPM does not systematically perform better than RCM for hydrological simulation because of some important biases on climate variables (here extreme precipitation). However, these results are consistent with some aspects of Xie et al. (2025) study.
• l447 : You should split the sentence at the comma and start again with “The study is conducted exemplarily…”
• l461 : From which study are you concluding of a “particular potential” ? This paper cannot conclude on a potential of ICON-CLM given the strong biases and the absence of added-value of using this CPM for hydrological simulations. I suggest you end your conclusion on an open note on the efforts put in place to correct some biases of the CPM (bias-correction, development, better understanding on extreme precipitation biases of ICON3km) to enhance CPM simulation and impact studies.

Bibliography

Arsenault, R., Brissette, F., Martel, J.-L., 2018. The hazards of split-sample validation in hydrological model calibration. J. Hydrol. 566, 346–362. https://doi.org/10.1016/j.jhydrol.2018.09.027 

Dale, M., Shelton, K., 2025. Convection-permitting models for managing hydrological extremes: practical, innovative examples. Philos. Trans. R. Soc. Math. Phys. Eng. Sci. 383, 20240289. https://doi.org/10.1098/rsta.2024.0289 

KLEMEŠ, V., 1986. Operational testing of hydrological simulation models. Hydrol. Sci. J. 31, 13–24. https://doi.org/10.1080/02626668609491024 

Shen, H., Tolson, B.A., Mai, J., 2022. Time to Update the Split-Sample Approach in Hydrological Model Calibration. Water Resour. Res. 58, e2021WR031523. https://doi.org/10.1029/2021WR031523 

Xie, K., Li, L., Chen, H., Xu, C.-Y., 2025. Assessing the performance of convection-permitting climate model in reproducing basin-scale hydrological extremes: A western Norway case study. J. Hydrol. 656, 132989. https://doi.org/10.1016/j.jhydrol.2025.132989

Review of 

“Do convection-permitting regional climate models have added value for

hydroclimatic simulations? A test case over small and medium-sized catchments in

Germany “

By Oakley Wagner, Verena Maleska, and Laurens M. Bouwer

This manuscript presents an evaluation of the convection-permitting regional climate model ICON-CLM 2.6.4 at 3 km resolution (ICON3km) compared to its driving model at 11 km with parameterized convection (ICON11km), focusing on the Weiße Elster basin in East Central Germany. The study assesses the ability of both models to reproduce key meteorological variables (air temperature, radiation, humidity, wind speed, and precipitation) and evaluates their suitability for hydrological impact modeling using the distributed hydrological model WaSiM. 

The analysis is performed for a 10-year period, from 2005 to 2014, focusing on verifying different atmospheric variables and  the capability of the atmospheric models to drive a hydrological model over a series of small or medium-size catchments. 

A comparison between the two  models and against observed data from different sources is carried on, with the final aim to address the potential added value of the convection-permitting model. 

The manuscript is well written and structured, presents new data and in general deserves to be published, although some revisions are needed to ensure that the results are better substantiated with figures/tables and that the conclusions are fully supported by the presented evidence.

General comments:

The abstract and introduction are clear, well-structured, and scientifically sound. They effectively present the study objectives, methods, and key results. 

Section 2. 

A more detailed discussion of the limited time period considered for a climatological analysis needs to be included in the manuscript, in particular in view of the fact that several results are characterized by statistically not significant differences between model’s results.

While subsection 2.2  (regarding the observational data is well described and comprehensive) , Section 2.3 (Climate data model) should be expanded. I would suggest adding some more information on the ICON model, of the model setup  and of the main parametrizations used. 

Apart from precipitation, the other atmospheric variables are verified against observations from the nearest ground station using the closest model grid value,  which can be particularly problematic for temperature. In complex terrain, the altitude of the model grid can differ significantly from that of the station. 

I believe it is worth adding some comments on this aspect in the subsection 2.4 and/or in the results section in the discussion of the biases, in particular for temperature.  

Section 3.

Overall, this section provides a useful comparison of ICON3km and ICON11km, but the analysis is uneven across variables. While the analysis of wind speed and precipitation is well-supported and convincing, the sections on air temperature, global radiation and relative humidity require additional figures.

Several  results are not supported by any figures/table, which in my opinion should be added at least as supplementary material. 

In Section 3.1.1 (Temperature), several detailed results are discussed (frequency distributions, diurnal cycle, seasonal variability of biases, DJF vs JJA differences). However, only the monthly mean biases are shown. The absence of a figure for the diurnal cycle or the frequency distribution makes it difficult for the reader to evaluate the stated findings and much of the text remains purely descriptive without visual support.

Similarly, in Section 3.1.2 (Global radiation)  the only figure presented relates to the diurnal cycle, claims about the frequency distribution of daily mean global radiation and the monthly bias (e.g., the July improvement of 2.5 J/cm² for ICON 3km) are not supported by any figure or table . Without such evidence, this part remains insufficiently substantiated.

As regards the diurnal cycle, Figure 3 shows a one-hour shift between the models and observations throughout the entire diurnal cycle (not only for the peak) and for all seasons. 

This perhaps deserves further investigation, or at least a verification of potential data misalignment, if this has not already been done

Section 3.1.3 (Relative Humidity), similar to the previous section regarding temperature and global radiation, describes  frequency distributions, monthly biases, and relative model performance (ICON3km vs ICON11km), but no figures are provided. As a result, the reader cannot verify whether the reported differences are meaningful or fall within observational uncertainty. 

Section 4

This part will probably need a slight revision after  the revision of section 3.

Section 5 

Overall, the conclusions are well written, clear, and consistent with the results presented in the manuscript. The authors provide a balanced discussion of both strengths (e.g., improvements in summer temperature, radiation, and wind speed representation) and limitations (notably the overestimation of heavy rainfall and its implications for discharge modelling).

Specific comments:

l145: The bandwidth of temperature is not clear. 

l150-153: The sentence is too long and could be split for clarity. 

l160-163: The sentence is too long and could be split for clarity.

l192: The sentence ‘Overall, for most months ICON 3km was found to outperform ..’ seems inconsistent with the preceding part of the paragraph which states that  “ ICON3km does not seem to offer noticeable improvement in the frequency distribution” (l190-191) and “ neither of the climate models shows significant difference to the observations. 

l247 : I suggest simplifying the phrase “in the summer month of July” to just “in July” . 

l445-448: The sentence is too long and could be split for clarity.